OPTIMIZING THE REPLACEMENT OF OVERHEAD LINES IN RURAL DISTRIBUTION SYSTEMS WITH RESPECT TO RELIABILITY AND CUSTOMER VALUE

Transcription

1 OPTIMIZING THE REPACEMENT OF OVERHEAD INES IN RURA DISTRIBUTION SYSTEMS WITH RESPECT TO REIABIITY AND CUSTOMER VAUE Patrk HIBER*, Bengt HÄGREN, na BERTING* *Royal Insttute of Technology (KTH), Karlstad Unversty - Sweden hlber@ets.kth.se SUMMARY In ths paper we present a method for establshng the value of a network s components from a relablty worth perspectve. The method can be appled to a general dstrbuton system.e. both for radal and meshed network systems. Moreover, ths paper shows results from an applcaton study for a rural network system that s domnated by overhead lnes. The purpose of the study was to establsh the value of a secondary feedng pont. Further, the method s used to establsh the best replacement strategy for the concerned overhead lnes. INTRODUCTION Customer nterrupton cost due to loss of supply s of crucal mportance for electrcal network owners. Crucal snce ths cost can be used as a measure of the relablty worth of a network [1]. One example of ths focus on customer nterrupton cost s the fact that the Swedsh Energy Agency, the government body that regulates network tarffs, wll apply a newly developed network performance assessment model for determnng the maxmum tarffs. In ths model, one of the most mportant factors s customer nterrupton cost. Ths fact combned wth the utltes progress toward cost effectve mantenance strateges calls for a new measure of component mportance, especally a measure that can be used for prortzaton of components based on where mantenance actons reduce customer nterrupton costs most effectvely. In [2] a method for dentfyng mportant components from a customer nterrupton cost perspectve s presented. The proposed mportance ndex, I H, and related mantenance potental render the possblty to evaluate network components from a system perspectve. Ths s acheved by a monetary nterpretaton of nterruptons, whch also enable further nvestgatons regardng mantenance actons and ther potental benefts. Ths paper utlzes the developed method n order to prortze the replacement of overhead lnes for a part of the rural electrc power network of Krstnehamn, Sweden. The goal s to prove the ndex useful and to dentfy areas of further development by means of an appled study. To demonstrate the suggested method we have studed the 11 kv dstrbuton network of Krstnehamns Energ Elnät, a local utlty wth 13 customers. The network covers the town of Krstnehamn ( the suburban network ) and the surroundng countrysde up to a radus of about 1 km ( the rural network ). The utlty has no producton of ts own, but draws power from the overlyng, regonal network through two recevng transformer statons. The purpose of ths paper s to answer the followng questons: What s the relablty worth of a secondary feedng pont for overhead lnes? Gven that the overhead lnes have to be replaced wth under ground cables, n whch order should the lnes be replaced? What s the mpact of the replacements n terms of nvestments costs and customer nterrupton costs? COMPONENT IMPORTANCE AND MAINTENANCE POTENTIA The utlzed mportance ndex s based on customer nterrupton costs. It s a partal dervatve of the total customer nterrupton cost, wth respect to the component s falure rate, [2]; I C H s = [SEK/f] (1) λ where I H s the mportance ndex, the studed component, C s [SEK/yr] the total expected customer nterrupton cost per year (where s denotes system.e. the whole network), and λ [f/yr] the falure rate of component. In order to calculate the expected customer nterrupton cost per year, C s, data s needed on; components (falure and repar rates), supply and load ponts, and network structure. Gven ths nput, a relablty analyss can be performed. The relablty analyss combned wth nterrupton cost data for the load ponts enables the calculaton of the expected total nterrupton cost. Next secton presents a descrpton of how the customer nterrupton cost was calculated n ths paper. H I corresponds to the total expected customer nterrupton cost (for all load ponts) that would occur f component faled. A hgh value of I H mples that a small change n falure rate for the component has a relatvely large mpact on the system avalablty (customer nterrupton cost). A low value mples that a change n falure rate does not affect the system as much. Consequently, the ndex can be used for prortzaton of components and/or mantenance actons. By multplyng each component s falure rate wth ts mportance ndex, I H, the component s mantenance potental can be estmated, (2). The component s mantenance potental can be nterpreted as the maxmum amount that can be saved by mprovng the relablty of the specfc component (.e. f the avalablty reaches 1%).

2 The mantenance potental and the mportance ndex can be used to dentfy the system s crtcal components, whch are dentfed on the bass of ther mpact on C s. C λ [SEK/yr] (2) H I APPICATION STUDY The suburban network of Krstnehamn conssts of underground cables and the rural network comprses seven separate overhead lne systems wth a total lne length of 11 km. Although the lnes are operated as radal, t s possble to close cross connectons between some of them durng dsturbances n order to shorten customer outage tmes. Ths paper shows result from applcaton studes for lne 1, one of the seven rural overhead lne systems. The average power consumpton for ths lne s 22 kw dstrbuted over 16 statons here referred to as load ponts; the total length s 12.1 km, mostly consstng of overhead lnes. Customer Interrupton Cost Customer nterrupton costs for the present study are based on a customer survey performed by Swedenergy [3]. The cost of an nterrupton consst of an ntal cost plus a cost that depends lnearly on the duraton. For every load pont nformaton about customer-composton and average energy consumpton s used to establsh the expected nterrupton cost. An expresson for the calculaton of the total expected nterrupton cost for the studed network, C s, s gven as follows; s = k + C (λ E c ) [SEK/yr] (3) where λ [nt/year] and E [kwh/year] are relablty ndces for load pont, k [SEK/nt] and c [SEK/kWh] are cost constants representng the customer types and composton and average power consumpton at load pont. Note that the load pont specfc λ and E are functons of nput data,.e. results from relablty calculatons based on falure rate, repar rate and network structure. Customers are dvded nto four groups as n Table 1. For example one load pont could consst of sx households (resdental) wth a total average power of 2 kw and one farm (agrcultural) wth an average of 1 kw. Ths would result n an ntal nterrupton cost of 14 SEK ( ) and a contnuously ncreasng cost of 43 SEK per hour ( ). For example 1.5 hours of nterrupton for that load pont results n a total customer nterrupton cost of 785 SEK ( ). TABE 1 - Interrupton costs [3] Type of customer SEK/kW SEK/kWh Resdental 2 4 Commercal ght ndustry 15 6 Agrcultural 1 35 Falure Rates The utlty keeps a complete record of nterruptons begnnng n 199. We have analyzed all nterruptons n the rural network durng the perod of , dvdng them up by falure cause. In Table 2 the causes of falures for overhead lnes are lsted for the studed 1 years. The number of falures s dvded by the total lne length (11 km) and the number of years to obtan the average falure rate,.12 falures/yr, km. The same approach s utlzed for the falure rates for load ponts and cables. The correspondng value for cables s.82 falures/yr, km. oad ponts (recevng statons) have a falure rate of.32 falures/yr. TABE 2 - Interrupton causes. Cause of nterrupton # falures Tree related 67 Weather related 26 Mscellaneous 16 Unclassfed 23 Sum 132 REIABIITY MODE The model of the network ncludes dsconnectors, cables, overhead lnes and statons, as shown n Fg. 1. In the analyss the dsconnectors are represented as connecton/dsconnecton ponts on the lne. Snce dsconnectors do not represent many falures of ths network they are not assgned any falure rate. Cables and overhead lnes are modeled wth a falure rate per klometer as stated n the prevous secton. Connectng networks are assumed to have 1% avalablty. Dstance [m] T16 T122 T121 T12 T11 T111 T123 T141 T17 T13 T131 T14 T132 T15 Dsconnector oad pont T151 S Dstance [m] Fg. 1. ne 1. The normal feedng pont s located n (,). The fgure s a representaton of the network and does not correspond to actual geographc postons, although lne lengths correspond to real lengths. The relablty model s based on the generalzed correctve mantenance routne descrbed below. A falure on a lne causes the crcut breaker at the supply pont of the lne to open. If the falure remans after a procedure wth automatc re-closures, we have an nterrupton. Generally two techncans are called upon, who locate and solate the faled

3 component wthn approxmately one hour. Ths solaton s acheved wth the dsconnectors and hence affects all downstream customers from the openng pont. After the falure s solated the power delvery s resumed to those customers who are stll connected. In the case of a secondary feedng pont, customers on both sde of the falure solaton area are provded electrc power. The remanng customers are reconnected when the fault s repared, whch takes approxmately two addtonal hours. The relablty model utlzed s an analytcal approxmatve model takng meshed structures nto account for the relablty calculatons. The modelng can be descrbed n three steps: 1. Identfy solaton segments. The segments are groups of components that can not be dsconnected from each other. All components wthn a segment are treated as serally connected, wth feedng ponts at the start of the segments and all load ponts at the end. 2. Identfy all possble falures, ther probabltes, and ther consequences on the whole network n terms of nterrupton tmes and C s. 3. Calculate relablty ndces, based on the prevously dentfed consequences and ther probabltes. In order to valdate the relablty model, we analyze t on system level (lne 1) wth the standard ndces SAIFI [falures/yr] (total number of customer nterruptons per year dvded by the total number of customers served) and SAIDI [h/customer, yr] (sum of customer nterrupton duratons per year dvded by the total number of customers). SAIFI and SAIDI are commonly used measures of system relablty for electrcal dstrbuton networks [1], [4]. There are also avalable methods usng these ndces n the process of fndng optmal strateges for dstrbuton plannng e.g. [5]. However, wth the utlzed falure model, the expected SAIFI becomes 1.8 falures/yr and SAIDI 2.3 h/customer, yr for lne 1. These expected results can be compared wth statstcs from the studed ten years ( ) that ndcate a SAIFI of.7 falures/yr and a SAIDI of 1.4 h/customer, yr for the actual lne. At frst t mght seem that the model s qute wrong,.e. only approxmately 4% of the predcted falures actually occurs. But t rather ndcates a possble weaknesses of the falure data, namely that falure rates are averages, and for example t s not accounted for whether a lne goes through forest or not. The dscrepancy for lne 1 can actually to a great extent be explaned by the fact that lne 1 manly goes through forest free areas, for example followng the ralroad and gong over felds. Based on ths new nformaton and Table 2, t would be reasonable to half the number of falures for lne 1. Dong ths, results n ndces that are very close to actual ndces,.e. expected SAIFI becomes.9 falures/yr and SAIDI 1.1 h/customer, yr. THE REIABIITY WORTH OF A SECONDARY FEEDING POINT In ths secton we analyze the relablty worth of a secondary feedng pont. In order to acheve ths we make an analyss of the two cases,.e. the radal case and the meshed case. Fgure 1 presents the modeled network (lne 1). In the fgure the secondary feedng pont s dentfed wth a damond. Radal Case Fgure 2 presents the computed mportance ndex (y-axs) plotted versus the mantenance potental (x-axs) for the radal case. The components close to the feedng ponts are, not surprsngly, the ones wth the hghest relablty mportance. The relablty mportance decreases wth the dstance to the supply pont or more specfcally wth the decreasng number of customers dependng on the functon of the component. One nterestng thng to note s the horzontal floor n the dagram (at 8 SEK/falure) that can be derved from the relablty model. That s, snce all components affect all customers untl the fault s dsconnected all components are mportant. In systems wth automatc swtchng devces avalable, ths floor generally does not exst, whch for example can be seen n [2]. ookng closely, t s possble to dentfy that the values n the plot are located on 11 horzontal lnes, that s 11 levels of relablty mportance. Ths, as well as the floor, derves from the relablty model and the functonalty of the network. Snce there are 11 areas that can be solated wth dsconnectors, all the components n these sectors wll be gven the same mportance. Ths s reasonable, snce the mportance of the components can be nterpreted as the nterrupton cost caused by the component n the case of falure. The sum of the mantenance potental for all components gves an estmate of the total expected nterrupton cost to 2 ksek/yr for the radal case. Importance ndex, I H [SEK/f] T15, T151 T16 T11 T122, T Mantenance potental [SEK/yr] Fg. 2. Radal case ndces, the two hghest and lowest levels of I H are marked wth dashed lnes, along wth ther respectve load ponts. Meshed Case Fgure 3 presents the mportance ndex plotted versus the mantenance potental for the case wth a secondary feedng pont (meshed case), utlzed n the case of nterrupton. One nterestng thng to note s that ths plot seems compressed compared to the correspondng plot for the radal case. The floor s stll the same, snce the utlzaton of the secondary feedng pont nvolves approxmately the same manual actons as the radal case. The exstence of 11 horzontal lnes s equally true for the meshed case as for the radal case. The

4 most mportant components are the ones that are located n the same solaton area as the load pont T11 followed by the components n the area wth T13 and T141. The reason for these components relatvely hgh mportance s that ther load ponts are hgh cost load ponts. The component wth the hghest mantenance potental s the lne connectng T13 and T14. Ths s one of the longest lne segments (components) and hence has a hgh falure rate. Combned wth a hgh relablty mportance, ths results n a hgh mantenance potental. Notce that snce ths s one of the longest lnes, t s also one of the most expensve to mantan/replace. The total expected nterrupton cost caused by lne 1 wth the meshed structure s approxmately 15 ksek/yr. Importance ndex, I H [SEK/f] T11 T122, T123 T13, T141 T12, T Mantenance potental [SEK/yr] Fg. 3. Meshed case ndces, the two hghest and lowest levels of I H are marked wth dashed lnes, along wth ther respectve load ponts. Evaluaton of the Secondary Feedng Pont The value of a secondary feedng pont can be establshed by comparng the radal case wth the meshed case. The analyss shows a reducton of customer nterrupton cost by approxmately 5 ksek/yr for the studed lne. Bearng n mnd that the extra feedng pont can also be used the other way,.e. feedng another lne (lne 2), the value of the secondary feedng could be roughly estmated to the double (1 ksek per year). Wth a net nterest of 6% over a 3-year perod ths corresponds to an nvestment of roughly 7 meters of cable, today. Snce lne 1 and 2 are separated wth a dstance of approxmately 7 meters, we can conclude that t was reasonable to connect them. PRIORITIZATION In the followng calculatons we wll consder the meshed case. Wth the mportance ndex presented n Fg. 3, t s possble to sort the components wth respect to ther relablty mportance nto a so-called prortzaton lst, presented n Table 3. The prortzaton lst can be used to dentfy components wth extreme relablty mportance (hgh and low), and specal actons can be consdered for these components. Components wth a hgh relablty value (I H ) should probably get much attenton n terms of mantenance and/or redesgn, whle components on the lower sde may be over-mantaned today. Another use of the relablty values s that they could be used to ndcate whether the current management of the network s approprate or f there are some unbalances. TABE 3 - Prortzaton lst for overhead lnes. oad ponts n the solaton area I H [SEK/falure] Overhead lne length [m] T T13, T T131, T T T T T T15, T T122, T T12, T The --- represent the small segment that not ncorporates any load pont. In the case of overhead lnes, a prortzaton lst provdes decson support for actons such as what segments to replace wth cables frst and where t s mportant to work wth ntense tree trmmng (as well as where t mght be wse to trm less ntensely). Hence, the prortzaton lst s a tool that can provde answer to the queston of n what order the overhead lnes should be replaced wth cables. Economcal Analyss In ths economcal analyss we focus on the replacement of overhead lnes wth cables. In the analyss the costs has been ncluded not only for replacement of the actual lne, but also for statons, 25 % dggng, a certan amount of BX and a number of mnor costs. These costs are estmated to averages per klometer n accordance wth a study performed by Swedenergy [6]. Note that costs for customer support and publc relatons are not ncluded. The major reason for not ncludng them n the analyss s that they are hard to estmate. Another dffcult to estmate cost, whch s not ncluded, s the admnstraton cost that occurs both n the case of preventve and correctve mantenance. Fgure 4, base case, dsplays the relatonshp between nvestment costs and customer nterrupton costs for overhead lnes, gven that they are replaced wth cables. nes wth hgh mportance ndces are replaced frst. Due to the relatvely low varaton n component mportance, the ponts n ths graph come close to a lne. The prevously mentoned 11 levels of mportance can be dentfed n ths pcture as 11 sets of slopes through the ponts. The reader may notce that the ponts n the rght part of the fgure are almost completely leveled out. Ths s due to the fact that these lnes actually already consst of cables and hence a replacement of them does not make sense (at least wth the utlzed assumpton of constant falure rate).

5 The dentfed relatonshp between customer nterrupton cost and nvestment costs clarfes the mpacts of nvestments n cables, gven that the nvestments are performed accordng to the prortzaton lst. Tree Scenaro The results of a smulaton based on that half of the lne segments go through forest can be seen n Fg. 4,.e. the tree scenaro. In ths scenaro we have assumed that overhead lnes n forest have a falure rate whch s 5% hgher than the average overhead lne, and that overhead lnes n open areas have a 5% lower falure rate. Ths s based on the approxmaton that 5% of all overhead lnes go through forest combned wth the nformaton gven n Table 2. Ths s an effect of that 67 out of 132 falures are tree related. As stated earler, lne 1 s not to a great extent affected by trees, but n order to evaluate the method we have made these assumptons. In Fg. 4 a dstnct levelng can be seen at approxmately 75 SEK of nvestment for the tree scenaro. Ths s where all overhead lne sectons that go through forest areas have been replaced wth cables. Hence, at ths pont further nvestments do not pay off as much as earler nvestments n terms of customer nterrupton costs. Due to the randomzaton of areas wth forest, the plot starts at a lower value than the prevous one. The cause of ths s that most of the forest-affected lnes are below average mportance. DISCUSSION The dfference between the hghest and lowest component mportance s not that large n ths case study. Ths results n that other factors mght contrbute more than the component mportance for mantenance prortzaton. One example s the status of wood poles,.e. one major reason for renewng mght be the condton of wood poles. Components n a generally bad condton are lkely to be prortzed, snce these most probably are prone to have more falures than the average component. The reason for why the presented method does not dentfy these crtcal components s the assumpton of a common falure rate (per km) for the overhead lnes. However, component condton mght be possble to nclude n the calculaton of mantenance potental by more detaled estmates of the falure rates. CONCUSION In ths paper we have presented a method for establshng the value of a network s components from a relablty worth perspectve. The method has then been appled to evaluate the value of a secondary feedng pont as well as the best replacement strategy for the nvolved overhead lnes. The concluson s that the proposed method can be appled to real lfe networks. It s data demandng but produces results that are valuable for the network owner. Customer Interrupton Cost [SEK/yr] Base case Tree scenaro REFERENCES [1] Bllnton, R. and Allan R. N., (1996). Relablty evaluaton of power systems. 2 nd ed. Plenum Press, New York. [2] Hlber, P. and Bertlng,., 24. Monetary mportance of component relablty n electrcal networks for mantenance optmzaton. Proc. Probablty Methods Appled to Power Systems 24. [3] Tapper, M. et al, 23. Electrc power nterrupton costs 23 (n Swedsh). Swedenergy Investment [SEK] x 1 6 Fg. 4. Total customer nterrupton cost as a functon of nvestment n underground cables, wthout contrbuton to nterrupton cost from statons. [4] Bertlng,. (22). Relablty Centred Mantenance for Electrc Power Dstrbuton Systems, KTH, Stockholm, ISBN: [5], F. and Brown, R. E., 24. A Cost-Effectve Approach of Prortzng Dstrbuton Mantenance Based on System Relablty. IEEE Trans. Power Delvery, vol 19. [6] Erksson Brger, 24, presentaton at Swedenergy.